Re: [Qemu-devel] Ping: [PATCH] qobject: object_property_add() performance improvement

[Peter Crosthwaite <peter.crosthwaite@xilinx.com>]

On Thu, Jun 4, 2015 at 12:35 AM, Pavel Fedin <p.fedin@samsung.com> wrote:
>  Hello Luiz! Have you missed this ?
>
> Kind regards,
> Pavel Fedin
> Expert Engineer
> Samsung Electronics Research center Russia
>
>
>> -----Original Message-----
>> From: qemu-devel-bounces+p.fedin=samsung.com@nongnu.org [mailto:qemu-devel-
>> bounces+p.fedin=samsung.com@nongnu.org] On Behalf Of Pavel Fedin
>> Sent: Thursday, May 28, 2015 11:42 AM
>> To: qemu-devel@nongnu.org
>> Cc: 'Luiz Capitulino'
>> Subject: [Qemu-devel] [PATCH] qobject: object_property_add() performance improvement
>>
>>  The function originally behaves very badly when adding properties with "[*]" suffix.
>> Nomnally these are used for numbering IRQ pins. In order to find the correct starting

"normally"

>> number the function started from zero and checked for duplicates. This took incredibly
>> long time with large number of CPUs because number of IRQ pins on some architectures
> (like
>> ARM GICv3) gets multiplied by number of CPUs.
>>  The solution is to add one more property which caches last used index so that
> duplication
>> check is not repeated thousands of times. Every time an array is expanded the index is
>> picked up from this cache. Cache property is a uint32_t and has the original name of the
>> array ('name[*]') for simplicity.
>>  Some more improvements:
>> - Call object_property_add_single() instead of recursing into itself - keeps off
> memcmp()
>> check every time when its result is already known.
>> - Allocate name_no_array only once and not every time for every property (there can be
>> thousands of them)
>>  The modification decreases qemu startup time with 32 ARMv8 CPUs by a factor of 2 (~10
> sec
>> vs ~20 sec).
>>
>> Signed-off-by: Pavel Fedin <p.fedin@samsung.com>
>> ---
>>  qom/object.c | 89 ++++++++++++++++++++++++++++++++++++++++++------------------
>>  1 file changed, 62 insertions(+), 27 deletions(-)
>>
>> diff --git a/qom/object.c b/qom/object.c
>> index b8dff43..72480bc 100644
>> --- a/qom/object.c
>> +++ b/qom/object.c
>> @@ -10,6 +10,8 @@
>>   * See the COPYING file in the top-level directory.
>>   */
>>
>> +#include <glib/gprintf.h>
>> +
>>  #include "qom/object.h"
>>  #include "qemu-common.h"
>>  #include "qapi/visitor.h"
>> @@ -721,35 +723,14 @@ void object_unref(Object *obj)
>>      }
>>  }
>>
>> -ObjectProperty *
>> -object_property_add(Object *obj, const char *name, const char *type,
>> -                    ObjectPropertyAccessor *get,
>> -                    ObjectPropertyAccessor *set,
>> -                    ObjectPropertyRelease *release,
>> -                    void *opaque, Error **errp)
>> +static ObjectProperty *
>> +object_property_add_single(Object *obj, const char *name, const char *type,
>> +                           ObjectPropertyAccessor *get,
>> +                           ObjectPropertyAccessor *set,
>> +                           ObjectPropertyRelease *release,
>> +                           void *opaque, Error **errp)
>>  {
>>      ObjectProperty *prop;
>> -    size_t name_len = strlen(name);
>> -
>> -    if (name_len >= 3 && !memcmp(name + name_len - 3, "[*]", 4)) {
>> -        int i;
>> -        ObjectProperty *ret;
>> -        char *name_no_array = g_strdup(name);
>> -
>> -        name_no_array[name_len - 3] = '\0';
>> -        for (i = 0; ; ++i) {
>> -            char *full_name = g_strdup_printf("%s[%d]", name_no_array, i);
>> -
>> -            ret = object_property_add(obj, full_name, type, get, set,
>> -                                      release, opaque, NULL);
>> -            g_free(full_name);
>> -            if (ret) {
>> -                break;
>> -            }
>> -        }
>> -        g_free(name_no_array);
>> -        return ret;
>> -    }
>>
>>      QTAILQ_FOREACH(prop, &obj->properties, node) {
>>          if (strcmp(prop->name, name) == 0) {
>> @@ -774,6 +755,60 @@ object_property_add(Object *obj, const char *name, const char
> *type,
>>      return prop;
>>  }
>>
>> +static void property_get_uint32_ptr(Object *obj, Visitor *v,
>> +                                   void *opaque, const char *name,
>> +                                   Error **errp);
>> +
>> +ObjectProperty *
>> +object_property_add(Object *obj, const char *name, const char *type,
>> +                    ObjectPropertyAccessor *get,
>> +                    ObjectPropertyAccessor *set,
>> +                    ObjectPropertyRelease *release,
>> +                    void *opaque, Error **errp)
>> +{
>> +    size_t name_len = strlen(name);
>> +
>> +    if (name_len >= 3 && !memcmp(name + name_len - 3, "[*]", 4)) {

Rather than if this large block, negate the condition and use
object_property_add_single to short return.

>> +        int i;
>> +        ObjectProperty *ret, *count;
>> +        /* 10 characters for maximum possible integer number */
>> +        char *name_no_array = g_malloc(name_len + 10);
>> +
>> +        count = object_property_find(obj, name, NULL);
>> +     if (count == NULL) {

Indentation doesn't look right.

>> +         void *v = g_malloc0(sizeof(uint32_t));
>> +

Can you allocate name_no_array here along with the count in a small
struct to save on some allocs and memcpys?

>> +            /* This is the same as object_property_add_uint32_ptr(), but:
>> +             * - Slightly faster and returns pointer
>> +             * - Will not recurse here so that we can use
>> +             *   raw name with [*] here */

newline before */

>> +            count = object_property_add_single(obj, name, "uint32",
>> +                                               property_get_uint32_ptr, NULL,

Do you need to register the getter or can you make it completely
opaque instead? Alternatively can you register the setter and use
set/get instead of going hands on with the property opaque pointer?

>> +                                               NULL, v, &error_abort);
>> +     }
>> +
>> +        name_len -= 2;
>> +        memcpy(name_no_array, name, name_len);

Can you drop the "[" as well? You will have one less byte to memcpy
and pick it up again in the sprintf so I dont see a huge perf
advantage. Makes the code below more readable.

>> +
>> +        for (i = *((uint32_t *)count->opaque); ; ++i) {

Parentheses not needed.


>> +         g_sprintf(&name_no_array[name_len], "%d]", i);
>> +

Indentation.

Regards,
Peter

>> +            ret = object_property_add_single(obj, name_no_array, type, get, set,
>> +                                             release, opaque, NULL);
>> +            if (ret) {
>> +                break;
>> +            }
>> +        }
>> +        *((uint32_t *)count->opaque) = i + 1;
>> +
>> +        g_free(name_no_array);
>> +        return ret;
>> +    }
>> +
>> +    return object_property_add_single(obj, name, type,
>> +                                      get, set, release, opaque, errp);
>> +}
>> +
>>  ObjectProperty *object_property_find(Object *obj, const char *name,
>>                                       Error **errp)
>>  {
>> --
>> 1.9.5.msysgit.0
>>
>
>
>